Background: This study examines whether risk factors related to incidence of depression are also related to prognosis, and whether a vulnerability-stress model can be established for prognosis. Methods: A prospective model for prognosis of depression (chronic or remitted course) in later life was studied in 236 depressed community-living elderly. Subjects were interviewed at baseline, and at follow-up 3 years later. Bivariate and multivariate relationships between risk factors and chronic depression (GMS-AGECAT) were assessed. Effect modification was studied between stressors and two types of vulnerability: vulnerability through a personal history of depression, and gender. Results: A personal history of depression, baseline functional limitations and incident anxiety syndrome predicted chronic depression, whereas life-events occurring between assessments, and changes in physical, functional or cognitive status did not. In subjects without a previous history, functional disabilities, male gender and receiving instrumental support correlated with a poor prognosis. The prognosis for subjects with a personal history of depression was not affected by other factors. In women, the development of chronicity was more strongly associated with a personal history than in men, whereas in men recent psychosocial and health-related characteristics were more important than in women. Limitations: Because the study consisted of two measurements with a 3-year interval, depressive episodes with a short duration may be under-represented. Conclusions: In the elderly, the impact of risk factors on the course of depression is modified by longstanding vulnerability characteristics, such as a personal history of depression and gender. More recent life stresses are related to prognosis in subjects without a personal history, and in men.